But improving overclocking means that they can increase the base clock of their chips from factory which would improve performance.I'd say that's very unlikely.
The cache architecture is a something that would have been decided ages ago, and changing it between Skylake and the Xeon E5/E7's (and Skylake-X) would only be done for a 'good' reason.
That 'good' reason is far more likely to be a response to where (at the time) Intel felt they were under pressure or needed to improve: server and HPC workloads.
Improving overclocking is hardly an important design criteria for Intel despite the impression which enthusiast forums might convey.